Research on Feature Extraction of CSP-based Channel Wave Signal

One of the necessary requirements of coal mine safety mining is to survey hidden disaster causing factors. Hidden disaster causing factors include geological structures such as scouring zones and faults. Channel wave seismic exploration is one of the most promising geophysical methods for exploring the interior of coal seams. In this paper, CSP is used to extract the features of channel wave signals and BP neural network is used to classify. The classification results are more than 95%, which proves its effectiveness. First, the channel wave signal is studied theoretically. Combined with COMSOL Multiphysics software to simulation modeling goaf, wash zone, collapse column and fault in coal measure strata are established. The Ricker wavelet is used as the source to transmit and the channel wave signal data set is received at the geophone. Then the spatial feature of the signal is extracted by CSP algorithm, and the extracted data is sent to BP neural network for training. The results show that this method can extract the characteristics of channel wave signals and the prediction error is small, which improves the accuracy of classification.